I was spending 8 hours a week on scheduling, client intake, and follow-ups. That's 400 hours a year of non-billable work.

So I built an AI assistant to handle it all.

Her name is Kiara. She runs 24/7 on my website, qualifies leads, books meetings, answers questions, and texts me when someone's serious. I haven't manually scheduled a call in months.

Here's exactly how I built her and why every consulting firm should have one.

The Problem Was Simple Math

Every new lead meant:

• 3-4 emails to find a meeting time
• 15 minutes updating my CRM
• Another 10 minutes on intake questions
• Follow-up emails when they went dark

Do that 10 times a week and you're burning a full day on admin.

For a solo consultant billing $250/hour, that's $100,000 in lost revenue annually. Just on scheduling.

The worst part? None of this work requires my expertise. It's pure operations.

Building Kiara: The Architecture

The setup is simpler than you'd think:

1. Client hits the chat widget on my site
2. Their message goes to Claude (Anthropic's AI)
3. Claude processes intent and context
4. Routes to the right action:
• Book a call → Opens my calendar
• Capture lead → Collects their info
• Answer question → Pulls from my knowledge base
• Send follow-up → Triggers email sequence

The magic is in the prompt engineering.

I gave Kiara a specific persona: professional but conversational, helpful but not pushy. She knows when to qualify a lead versus when to just answer a question. She remembers context from earlier in the conversation.

Most importantly, she knows when to tap out and loop me in.

Here's a real exchange from last week:

Visitor: Do you work with SaaS companies?
Kiara: Yes, we work with several SaaS companies between $1M-$20M in revenue. What specific operational challenges are you facing?
Visitor: We're struggling with customer onboarding. Takes too long and we're losing deals.
Kiara: Customer onboarding is a common bottleneck we help optimize. Derek typically reduces onboarding time by 40-60% through process mapping and automation. Would you like to discuss your specific situation? I can show you his calendar.

She qualified the lead, demonstrated expertise, and offered next steps. All while I was in a client meeting.

The Logic Layers That Make It Work

Four layers keep Kiara from going off the rails:

1. Intent Detection
Every message gets classified: scheduling request, service question, pricing inquiry, or general chat. This happens before any response.

2. Context Memory
She tracks the full conversation history. If someone mentions their company name once, she remembers it throughout.

3. Action Routing
Based on intent + context, she picks the right path. Scheduling goes to Calendly. Serious leads get my intake form. Tire kickers get helpful content.

4. Fallback Handling
When she's unsure, she asks clarifying questions. When she's really unsure, she says "Let me connect you with Derek directly" and texts me.

The whole system runs on guardrails. She can't quote prices. Can't make promises. Can't discuss specific client work. Just handles the operational basics.

The Results After 6 Months

The numbers tell the story:

• Zero back-and-forth scheduling emails
• 73% of leads fully qualified before I talk to them
• 8 hours per week reclaimed
• 24/7 response time (huge for international clients)

But the real win is psychological. I'm not context-switching between client work and admin tasks anymore. When I'm consulting, I'm fully present. When a lead is ready, they're actually ready.

Is it perfect? No.

She still struggles with complex questions. Sometimes misreads intent. Once told someone I specialized in "strategic banana operations" (I'd been testing fruit vendor examples).

But she gets better every week. Each conversation teaches the system. Each edge case becomes a learning opportunity.

This is what AI should be doing for small firms. Not replacing the consultant — handling the operations so the consultant can consult.

Most firms under $20M can't afford a full-time ops person. But they can build an AI agent in a weekend that handles 80% of the repetitive work.

The technology is here. The question is whether you'll use it or keep trading billable hours for scheduling emails.

Your expertise is what clients pay for. Everything else is just overhead.